DOT Project Number:  90-00-LRTF-612

Fiscal Year:  2006

Award:  $9,000

Principal Investigator:  Dr. Brian Wilsey, Department of Ecology, Evolution and Organismal Biology, Iowa State University, bwilsey@iastate.edu

Summary Report:

DOMINANT NATIVE PRAIRIE-GRASS SPECIES DIFFER IN THEIR BIOMASS PRODUCTION AND SUPPRESSION OF SUBORDINATE SPECIES

Introduction

Many prairie communities are strongly dominated by grass species and the abundance of these species can suppress the establishment of rare forb species to reduce species diversity (Baer et al. 2005, Williams et al. 2007).  The proportion of biomass production from C₄ (warm season) grasses within and among prairies can vary greatly (Martin et al. 2005), but it usually makes up a substantial portion of any given plot (Turner and Knapp 1996, Wilsey and Polley 2003).  However, plant species diversity is dominated by the proportional composition of forbs (Turner and Knapp 1996), and diversity is what most people are concerned with when they restore prairie (e.g. Palmer et al. 1997).  Researchers at Konza Prairie in Kansas have found that the dominant C₄ grass big bluestem, Andropogon gerardii can greatly suppress plant diversity in the Flint Hills prairie region.  When dominance of Andropogon gerardii is reduced by grazing or a reduced fire frequency, the diversity of forbs and cool season grasses (and indeed the whole plant community) increases (Hartnett et al. 1997, Collins et al. 1998).  In a more mesic system, Williams et al. (2007) found that frequent mowing of C₄ grasses led to higher forb establishment in a grass-dominated planting. This suggests that dominance by grasses has the potential to regulate the diversity of restorations, and further study of what causes grass dominance is warranted. 

Many restoration projects are currently being planted with cultivars (Jones 2003), and new perennial biofuel plantings will be based on using cultivars in most cases.  In addition to concerns about the possibility of cultivars hybridizing with remnant individuals (Lesica and Allendorf 1999, Gustafson et al. 2002), the dominance by these cultivars on other species in mixture may be higher than what would be found for locally collected genotypes.  Cultivars are usually selected for high seed germination rates and increased ‘vigor’ and production, but whether these traits are enhanced over local genotypes, and whether these traits are important to the ecology of developing prairies is largely unknown or undocumented.  Huston (1994) hypothesized that the highest plant species diversity occurs with both intermediate amounts of disturbance and low growth rates of constituent species.  Low growth rate is predicted to increase diversity by limiting the rate of competitive exclusion.  Since cultivars are usually selected for rapid growth rate (high “vigor”) (Gustafson et al. 2004), then primary productivity may be higher but forb recruitment and species diversity might be lower in plots dominated by these species compared to plots dominated by slower growing native genotypes.  Although cultivar status was not the focus of their restoration study, Baer et al. (2005) found that a cultivar of the lowland species Panicum virgatum attained very high dominance and suppressed local diversity.   If cultivars do indeed dominate plots more than locally collected plants, than management objectives of high productivity and high species diversity would be in conflict if cultivars are used.  On the other hand, if there is no difference in diversity between projects planted with cultivars and native genotypes, then some managers may opt to use cultivars because more acreage can be planted with the cheaper, more readily available, cultivar seed (Jones 2003).  In the present study, we are comparing the development of prairie plantings that are dominated by multiple C₄ grass species, either collected from local or from cultivar seed sources.  Thus, we are testing Huston’s model without species differences to confound high vs. low growth rate comparisons. 

I suggest that basic ecological and evolutionary theory (reviewed by Lesica and Allendorf 1999) predicts three possible outcomes for studies that compare native and cultivar seed in planted prairies (Figure 1).  The cultivar vigor hypothesis predicts that human selection for increased vigor will lead to increased resource capture and aboveground biomass production in cultivars compared to locally collected genotypes.  In this scenario, cultivar-planted prairies would have more productive grasses and a lower recruitment of other native species (e.g. forbs).  Conversely, the local adaptation hypothesis predicts that cultivars will capture fewer resources and will be less productive than locally-collected genotypes.  This is because the original cultivar seed was typically collected from a more distant location than local seed.  If local adaption is especially prevalent and strong, then cultivar genotypes would be less productive regardless of any human selection for increased vigor.  Both hypotheses received partial support by Gustafson et al. (2004): the Rountree cultivar of Andropogon gerardii had higher biomass and heights than did plants from local seed and plants from a distant remnant source had lower biomass than plants from local seed.  However, a second cultivar (Pawnee) did not differ from plants from local seed sources.  A final hypothesis is the null hypothesis, which predicts that there will be little or no differences between cultivars and local genotypes.  This is possible if the two processes (human selection for increased vigor, local adaptation) cancel each other out.

The cultivar vigor and local adaptation hypotheses have important predictions for relationships between species diversity and productivity in grassland plantings.  If cultivars were human-selected to be vigorous and to have high germination rates, then they might have greater interspecific-intraspecific competition ratios compared to locally collected genotypes.  This destabilizing effect (Chesson 2000) might lead to greater declines in diversity over time in cultivar-dominated than in non-cultivar-dominated grasslands.  In either case (i.e. cultivars or local genotypes), ecological theory predicts that productivity will be higher in mixtures if species utilize resources differently in time or space (i.e. have greater niche partitioning) (Tilman et al. 1997).  Dominant grasses in central and southern tallgrass prairies are functionally similar in that they all are all C₄ grasses.  However, there are large functional differences in growth form (e.g. rhizamotous vs. bunchgrass) and heights among these species, and these differences may lead to increased biomass production in mixtures due  to having higher resource uptake in space or time.  Silletti and Knapp (2002) found that Andropogon gerardii and Sorghastrum nutans responded differently to climatic variables and fire frequency.  If the functional differences seen among C₄ grasses are important, then we would predict that productivity in mixtures will be higher on average than productivity in their corresponding monocultures.  These differences are predicted to be larger in grassland plantings dominated by locally collected seed than in plantings dominated by cultivars.

Here I test these ideas by comparing biomass production, subordinate species recruitment, and weed suppression among grassland plots dominated by different grass species in the Loess Hills of western Iowa.  At the neighborhood and patch scales, prairie ecosystems in this area can be dominated by a variety of C₄ (warm season) grasses in addition to A. gerardii, including Sorghastrum nutans (indian grass), Schizachyrium scoparium (little bluestem) or Bouteloua curtipendula (side-oats grama) in upland locations (Brudvig et al. 2007), and Panicum virgatum (switchgrass) in lower areas (Novecek et al. 1985).  I report on how different warm season grasses vary in their biomass production, weed suppression, and subordinate prairie species recruitment.

Methods

Study Site and Field preparation

The study was conducted on Iowa State University owned lands in the loess hills region of Iowa in Monona County (Western Research Farm).  The official weather station on site receives an average of 762 mm of precipitation per year.  The soils are very deep silty loess soils.  Experimental plots were located on a hill-top in a 16 ha abandoned pasture formerly dominated by smooth brome Bromus inermis.  The area was grazed by cattle until 2002 and was not fertilized for many years.  The field was prepared by disking three areas (blocks) during fall 2004 and again in early spring 2005 just prior to planting.  The 2005 growing season had precipitation (658 mm) that was slightly below the 30 year mean with a wetter than normal April, May, and June and drier than normal July and August.

Experimental Design

The experiment consisted of planting equal-sized seedlings of one of five native grass species (Andropogon gerardii, Sorghastrum nutans, Panicum virgatum, Schizachyrium scoparium, or Bouteloua curtipendula), mixtures of all five species, or no grasses at all into experimental plots during early May 2005 in a randomized block design.  These treatments were crossed with seed-source treatments, with seedlings being either from remnant collected seed (Custom Seed Co., Walnut, IA) or from cultivars.   Treatments were randomly assigned to plots in a 6 (each of the five species in monoculture plus mixtures of all five species) x 2 (plants from local or from cultivar seed) factorial design within three blocks (southwest-, north-, or east-facing slope).  There were 2 replicate monocultures within each block for 5 species x 2 seed source x 3 blocks x 2 reps = 60 monoculture plots total.  There were 4 replicate mixture plots within each block for a total of 2 seed source x 3 blocks x 4 reps = 24 mixtures total.  Twelve companion bare ground plots (4 within each block) were also included to test if subordinate and weed species establishment would be greater in grass-free plots.

Transplants were used instead of seed to control the rate of establishment and plant density, which enables more careful comparisons across species.  A previous attempt at testing these hypotheses involved seeding bare ground in 5 x 5 m plots.  However, these plots had very uneven establishment among species treatments and very high amounts of weed invasion in all plots and had to be abandoned.  In order to prevent this differential establishment among species during the seedling recruitment stage, we planted equal-sized seedlings under controlled densities in initially weeded plots on a common soil type.  Because the study was so labor intensive and controlled, we had to use relatively small plots.   Seedlings were planted in each 1 m² plot at a density of 72 plants per plot.  As a result, this study is most relevant to understanding local, neighborhood-scale processes and less relevant to understanding larger scale processes such as spatial patchiness and other processes that affect larger-scale diversity.

Plots were hand watered for one week to facilitate establishment of grasses and were weeded until the grass canopy had established (i.e. until July 13, 2005).  Thereafter, weeds were allowed to freely colonize and grown in the plots.  Grass transplant survival rate was greater than 95 % in all plots. Alleyways between plots were mowed monthly during the duration of the study.

Sampling of Plant Traits

Seed germination rate was estimated among species and cultivars in two trials using seeds of each of the C₄ grass species and cultivars.  Each of the two trials had 3 replicates per species per trial.  Trials were conducted in field-collected-soil in well watered pots (50 seeds per pot) in a greenhouse at Iowa State University. 

Developing communities such as prairie restorations often begin with very open canopies on bare soil, as was the case here.  Dominant grasses in the different treatments were predicted to differentially fill-in space during canopy development in the three dimensions of space.  To test this, estimates were made of traits associated with resource capture to determine whether grasses differed across species and between locally collected and cultivar plants within each monoculture plot.   Measurements were made on easily measured traits associated with total resource uptake (light uptake and total percent cover as proxies for total resource capture) and both upward (height) and lateral spreading (length and width).

Canopy light penetration and percent vegetation cover was measured in each plot in July and September, 2005.  Canopy light capture was estimated by placing a 1-m Decagon (Pullman, WA) ceptometer and comparing light above and below the plant canopy during mid-day (10:00 to 2:00 p.m. standard time).  The ceptometer was placed diagonally into each plot in two locations (NW to SE and NE to SW) below the canopy at the soil surface.  The end of the light bar was always at least 10 cm from the corner of the plot.  Soil-surface light values were compared to light values above the canopy (below/above) to estimate the proportion of light that reached the soil surface, and this value was subtracted from one for estimates of capture.  Percent vegetation cover was separately visually estimated separately in each of the four quarters of each plot (i.e. for each 0.25 m²).  This was done to improve the accuracy of plot-level estimates by sampling a smaller area.  These values were then averaged across the four estimates per plot to obtain one cover estimate per plot.  Small sheets of calibration paper of known cover of 0.1, 1.0, 5 and 10% were used to initially calibrate the cover estimates, and all estimates of cover were done by the same person to reduce observer bias.

Plant height and basal area were measured in the first year of establishment (2005).  Height was measured from the soil surface to the base of the upper-most leaf on three plants per plot.  The basal area of each plant approximated a circle.  Therefore, basal area was estimated by measuring two plant diameters between the farthest tillers at the base of three plants per plot.  These values were then converted into one estimate of basal area per plot with the standard equation for the area of a circle (area = πr²) using the mean radius of the two measurements.  For each variable, measurements were averaged across the three plants per plot to prevent pseudoreplication.

Peak aboveground biomass was harvested as an estimate of aboveground net primary productivity in the second growing season.   Aboveground peak biomass was estimated by clipping biomass to 2 cm on September 22-23, 2006.  Live material was sorted by species, dried at 65 C 48 hours until dry, and weighed.

Seed Additions of subordinate species

Subordinate species, which were mostly prairie forbs (Table 1), were added to the plots in a seed  mix after grasses had established.  These species were added to test how grass treatments would suppress diversity through suppressing establishment of subordinate species.  These species are key to having high species diversity, and are potentially important in the long run to prairie persistence due to their N-fixing ability (leguminous forbs) and their weed suppression abilities (non-leguminous forbs, e.g. Losure et al. 2007).  Seeds from 26 native prairie species were added to each plot on June 15 and December 16, 2005 (Table 1).

Statistical Analysis

Peak biomass variables were total biomass (grass + weeds + subordinate species from the seed mix), weed biomass alone, and subordinate species biomass alone.  These variables were analyzed with randomized block ANOVA to test for dominant species effects (6 levels), cultivar effects (2 levels), and their interaction.  Block by treatment interactions were pooled into the error term a priori (as is the standard practice, Peterson 1985).  Main effect differences among species were tested with Tukey’s post-ANOVA test.  The species*cultivar interaction was further tested with the SLICE option (Littel et al. 2002).  The SLICE option tested cultivar vs. non-cultivars for each species when the interaction was significant (P < 0.05).  Germination rates were analyzed with a similar approach and model, except that blocking was done on trial.

Resource capture trait data were analyzed first with principal components analysis to test for whether variables were independence.  There were two major principal components of variation in the data (i.e. two axes with eigenvalues > 1.0).  Light capture (0.62), percentage cover (0.61), and height (0.48) all loaded heavily on axis 1, which accounted for 54.4% of the variation in the data.  Basal area had a low loading of 0.10 on axis 1.  Axis 2 was explained by a trade-off between basal area, with a loading of 0.86, and height, which had a loading of -0.47.  Loadings of other variables were < 0.22.  Axis 2 accounted for 29.8% of the variation.  Because height (axis 1) and basal area (axis 2) were largely independent (univariate correlation of -0.22), I analyzed how these two variables varied among treatments with univariate ANOVA’s.  These two variables were then regressed against light capture and percent cover to determine if they were related to overall resource capture.

Biomass variables were compared between the bare ground and vegetated plots with a Dunnets test in a one-way ANOVA.  Dunnet’s test compares a control, in this case the grass-free plots, to each of the other 12 planted treatments in turn while controlling the type 1 error rate. 

Results

Plant traits

Germination rates varied significantly among species (F_1,49 = 47.0, P < 0.0001) and were different between cultivars and local genotypes in every species pair (Cultivar F_1,49 = 44.7, slice by species, all P values < 0.01).  Cultivars had higher germination rates in general than did local genotypes with differences ranging from a 32-fold higher germination rate in S. nutans to a 1-fold higher rate in S. scoparium cultivars (Table 2).  However, there was an exception to the rule in that A. gerardii non-cultivars had 6-fold higher germination than cultivars (Species x Genotype interaction, F_4,49 = 61.0, P values for each species pair < 0.0001, Slice P < 0.0001).

There were very large differences in height among species (species main effect,  P < 0.01), but there was no simple difference in height between cultivars and non-cultivars (main effect, F_1,47= 2.9, P = 0.095).  Switchgrass, big bluestem and indian grass were much taller than little bluestem and side-oats grama (Figure 2).  Height was different between cultivars and non-cultivars in 3 out of 5 cases, but the difference varied among species (interaction, P<0.01) and with time (P < 0.01).  Cultivars were shorter than non-cultivars for indian grass (time 1, P < 0.1) and switchgrass (time 2, P < 0.01), whereas cultivars were significantly taller in little bluestem (time 1, P = 0.04). 

Basal area varied among species in a manner that was independent of heights (Figure 3).  The shortest species side-oats grama had the greatest basal area (species main effect, F_4,47 = 4.87, P < 0.01) and this difference between side-oats grama and other species increased over time (species x time, P < 0.01).  Side-oats grama had significantly greater basal area than indian grass in the first time period, and it had greater basal area than switchgrass and big bluestem in the second period.   Cultivars had 18-19 % wider bases than non-cultivars (cultivar main effect, F_1,47 = 4.44, P = 0.04), and this difference was consistent over time periods (Figure 4).

Canopy light capture, which serves as a proxy variable for total resource capture, was positively related to height and basal area during the early (July) sampling period (height slope=0.012, area slope=0.009, combined r²=0.36, P<0.01 for both variables), but it was only related to height during the later (September) sampling date (height slope = 0.009, r²=0.15, P<0.01).

Aboveground net primary productivity (peak biomass)

There were significant differences among dominant grass treatments in their biomass production (Figure 5).  Indian grass, switchgrass, and little bluestem were more productive on average (mean across species of 661.5 g/m²) than big bluestem or side-oats grama (ANOVA, Duncan's tests, P values < 0.05), which averaged 424.1 g/m².  Differences in lateral spread between seedlings from locally collected seed and cultivar seed did not result in greater productivity: there was no significant difference in productivity between plots planted with seedlings from locally collected seed and cultivar seed (P > 0.05).  There was also no difference in productivity between single species plantings and five-species mixtures (P > 0.05).  The overall mean for monocultures was 563.5 g/m² vs. a mean of 566.5 g/m² for mixtures. Not surprisingly, biomass production was much higher in every planted grass treatment than it was in unplanted plots (Dunnets test, difference between unplanted and all planted treatments P < 0.05).

Weeds (non-planted or seeded species) generally made up less than 10 % of the total biomass at harvest in planted plots (Figure 5).  Nevertheless, there were significant differences among species.  Little bluestem had more weed biomass at 54.1 g/m² than did switchgrass at 13.1 g/m² when averaged across cultivar-non cultivar groups.  Bare ground plots had between 5X (little bluestem non-cultivars at 56.4 vs. 263.5 g/m2) and 35X higher weed biomass (switchgrass cultivars at 7.6 vs. 263.5 g/m² in bare ground plots) than planted grass plots (Dunnets test, all P values < 0.05).

Subordinate species establishment

Seeded species biomass was dominated by the early successional species Verbena stricta, which made up the majority of seeded biomass.  Biomass of seeded species varied significantly among the species treatments.  Little bluestem had higher seeded species biomass (44.7 g/m²) than did other species (5.2 - 14.9 g/m2) (ANOVA and Tukey's test, P values < 0.05).  This suggests that early prairie forb establishment was higher in plantings that contain little bluestem than plantings containing other species.

In general, unplanted plots did not have greater amounts of seeded species than planted plots.  In only one case, unplanted vs. Panicum virgatum cultivar plantings, was there are significant difference in seeded species biomass (Dunnets test, P < 0.05, all other comparisons non-significant), with P. virgatum cultivars having less seeded species biomass than the unplanted controls.  In each of the other 11 cases, there was no difference in seeded species biomass between planted and unplanted control plots.

Discussion and Conclusions

There are many prairie restoration projects ongoing and getting started that are addressing how to increase species diversity in grassland systems (e.g. Mlot 1990, Smith 1998, Hötxel and Otte 2003, van Diggelen and Marrs 2003, Prach 2003).  However, projects are usually hampered by a lack of knowledge on how to restore the high diversity found in intact native grasslands, and the species diversity of plantings is often much lower than species richness of native prairie remnants (Kindscher and Tieszen 1998, Sluis 2002, Martin et al. 2005).  Further information is needed on how to create the combinations of species and environmental conditions necessary to successfully establish a diverse prairie (Howe 1994, Blumenthal et al. 2003).

Here, we found that C₄ grass species identity, but not species richness or seed source, affected productivity and subordinate species establishment.  Cultivars differed from non-cultivars in their heights, but in some cases the cultivar was taller and in some cases the cultivar was shorter than plants from local seed sources.  Basal area was more consistently higher in cultivars.  These differences suggest that cultivars are genetically different than non-cultivars. Increased basal area could be associated with human selection for increased vigor.  However, these differences between cultivars and non-cultivars did not result in differences in biomass production or subordinate species establishment.  Biomass production and subordinate species establishment were more associated with differences among grass species.

The high frequency of significant species x cultivar interactions on traits, and the lack of a significant cultivar effect on light capture, biomass production, or subordinate species recruitment suggests that both processes underlying the cultivar vigor and the local adaptation hypotheses are operating, but that they may be cancelling each other out.  For most species, cultivars had traits that appeared to make them more vigorous, for example higher basal area and higher seed germination.  In a few cases, cultivars were taller than non-cultivars.  However, cultivars were found to be shorter in other cases.  The fact that the cultivars tend to not be adapted to the area could have caused a corresponding reduction on biomass (reduced fitness) that counteracted the human selection for increased vigor.   This counteracting balance between these two forces (increased vigor and non-local adaptation) could have prevented significant differences from occurring between cultivars and non-cultivars. 

One caveat to keep in mind when interpreting these results is that we used transplants in this study.  By transplanting seedlings, we bypassed the seedling establishment phase.  The seedling establishment phase is a critical one.  Thus, the differences found in seedling emergence between cultivars and non-cultivars could be important during early stages of prairie establishment, and this deserves further study with seeded plots.

Some have suggested that having no C₄ grasses in the seed mix, at least initially, would lead to increased forb recruitment.  C₄ grasses could then be seeded in later years after forbs have established.  This approach would work only if weed invasion is either very low in abundance and biomass or weed abundance and biomass are irrelevant to prairie species establishment.  A previous restoration study at this same site had to be abandoned because weeds prevented prairie establishment.   These seeded plots were heavily dominated by Chenopodium album (lambs quarters) in year one and Coronilla varia (crown vetch) and Bromis inermis (smooth brome) in years two and three.  The results from this earlier study suggest that excessive weed establishment can cause projects to fail.  In the current study, plots had much higher weed biomass in bare ground plots where grasses were not planted.  Furthermore, they did not have higher abundance or biomass of species from the seed mix.  Although this study is still in its early stages, the results so far suggest that having no grasses planted initially is not better than having grasses planted. 

Highest establishment from the seed mix and highest light at the soil surface occurred in plantings of little bluestem.  These plantings had much less weed biomass, but the same amount of biomass from species in the seed mix (mostly the disturbance favored Verbena stricta so far) as bare ground plots.   Based on this, I suggest that seed mixes using only little bluestem as the C₄ grass are most likely to achieve the objectives of having higher forb recruitment while keeping weeds to an acceptable minimum.

Results to date suggest that native perennial prairie grass species differ in their biomass production.  Switchgrass, indian grass, and little bluestem were highly productive.  Little bluestem also contained the highest amount of prairie forb biomass.  Big bluestem and side-oats grama were less productive than other species at this site.  Growing species in mixture did not increase or decrease productivity over monocultures, and did not have significant different effects on weed or subordinate species establishment.  All dominant grass species used were C₄ grasses, which suggests that they are similar functionally in when they grow, flower and go dormant (i.e. similar phenology).  However, they did have very different heights and growth forms, ranging from tall rhizomatous species to short bunch-grass species.  These differences did affect how they filled in space during canopy development.  Taken together the lack of a difference between mixtures and monocultures suggests that biomass production may not be higher in mixtures than monocultures when legumes and other forbs are not in the mix or when insufficient phenologically related functional diversity is present. 

At the same field site as this, Losure et al. (2007) and Isbell et al. (submitted) found no change in productivity or invasion resistance across plots that varied in their among-plant height variability.  However, both variables increased with the proportion of early emerging forbs in the mix.  This suggests that mixtures can have attributes such as greater pest resistance (Kennedy et al. 1998, Wilsey and Polley 2002, Losure et al. 2007) or productivity if plantings include plant species that grow at different times of the year.  This makes mixtures attractive candidates for grassland plantings in most situations.  I suggest that mixed plantings with little bluestem will facilitate the recruitment of both the leguminous and non-leguminous forbs that provide the greatest benefit to prairie persistance.  

Acknowledgments

Thanks to Kim Wahl, Andrea Blong, Leanne Martin, Anna Loan-Wilsey, Wayne Roush and Don Hummel for all of their help in the field.  This project was funded by the Iowa Department of Transportation Living Roadway Trust Fund and Iowa State University.

Literature Cited

Baer, S.G., J.M. Blair, S.L. Collins, and A.K. Knapp.  2003.  Soil resources regulate productivity and diversity in newly established tallgrass prairie.  Ecology  84:724-736

Baer, S.G., S.L. Collins, J.M. Blair, A.K. Knapp, and A.K. Fiedler.  2005.  Soil heterogeneity effects on tallgrass prairie community heterogeneity: an application of ecological theory to restoration ecology.  Restoration Ecology 13:413-424.

Blumenthal, D.M., Jordan, N.R., and M.P. Russelle.  2003.  Soil carbon addition controls weeds and facilitates prairie restoration.  Ecological Applications  13:605-616

Brudvig, L.A., C.M. Mabry, J.R. Miller, and T.A. Walker.  2007.  Evaluation of central North American prairie management based on species diversity, life form, and individual species metrics.  Conservation Biology 21:864-874

Collins, S.L, Knapp, A.K., Briggs, JM, Blair, JM and E.M. Steinauer.  1998.  Modulation of diversity by grazing and mowing in native tall-grass prairie.  Science  280:745-747.

Gustafson, D.J., Gibson, D.J. and D.L. Nickrent.  2004.  Competitive relationships of Andropogon gerardii (Big bluestem) from remnant and restored native populations and select cultivated varieties.  Functional Ecology 18:451-457

Hartnett, D.C., Hickman, K.R. and L.E. Fischer Walter.  1996.  Effects of bison grazing, fire, and topography on floristic diversity in tallgrass prairie.  Journal of Range Management  49:413-420.

Howe, H.F.  1994.  Managing species diversity in tallgrass prairie:  assumptions and implications.  Conservation Biology  8:691-704.

Huston, M.A.  1994.  Biological Diversity.  Cambridge University Press, Cambridge, UK.

Jones, T.A.  2003.  The restoration gene pool concept: beyond the native versus non-native debate.  Restoration Ecology  11:281-290.

Kennedy, T.A., S. Naeem, K.M. Howe, J.M.H. Knops, D. Tilman, and P. Reich.  2002.  Biodiversity as a barrier to ecological invasion.  Nature  417:636-638

Kindscher, K. and L.L. Tieszen.  1998.  Floristic and soil organic matter changes after five and thirty-five years of native tallgrass prairie restoration.  Restoration Ecology  6:181-196.

Lesica P. and Allendorf, F.W.  1999.  Ecological genetics and the restoration of plant communities: mix or match?  Restoration Ecology 7:42-50

Losure, D.A., Wilsey, B.J. and K.A. Moloney.  2007.  Evenness-invasibility relationships differ between two extinction scenarios in tallgrass prairie.  Oikos 116:87-98

Martin, L.M. and B.J. Wilsey.  2006.  Assessing grassland restoration success: relative roles of seed additions and native ungulate activities.  Journal of Applied Ecology 43:1098-1110

Martin, L.M., Moloney, K.A. and B.J. Wilsey.  2005.  An assessment of grassland restoration success using species diversity components.  Journal of Applied Ecology 42, 327-336

Mlot, C.  1990.  Restoring prairie.  Bioscience 40:804-809.

Novecek, J.M., D.M. Roosa, and W.P. Pusateri.  1985.  The vegetation of the Loess Hills landform along the Missouri River.  Proc. Iowa Acad. Sci.  92:199-212.

Palmer, M.A., Ambrose, R.F. and N.L. Poff.  1997.  Ecological theory and community restoration ecology.  Restoration Ecology  5:291-300.

Polley, H.W., Wilsey, B.J., Derner, J.D., Johnson, H.B. and J. Sanabria.  2006.  Early-successional plants regulate grassland productivity and species composition: a removal experiment.  Oikos 113:287-295.

Polley, H.W., Wilsey, B.J. and J.D. Derner.  2005.  Patterns of plant species diversity in remnant and restored tallgrass prairies.  Restoration Ecology 13:480-487

Polley, H.W., Wilsey, B.J., Derner, J.D., Johnson, H.B. and J. Sanabria.  2006.  Early-successional plants regulate grassland productivity and species composition: a removal experiment.  Oikos in press

Smith, D.  1998.  Iowa Prairie: original extent and loss, preservation and recovery attempts.  Journal of the Iowa Academy of Sciences  105:94-108.

Silletti, A. and A. Knapp.  2002.  Long-term responses of the grassland co-dominants Andropogon gerardii and Sorghastrum nutans to changes in climate and management.  Plant Ecology 163:15-22.

Sluis, W.J.  2002.  Patterns of species richness and composition in re-created grassland.  Restoration Ecology  10:677-684.

Tilman, D., Lehman, C.L., and K.T. Thomson.  1997.  Plant diversity and ecosystem productivity: theoretical considerations.  Proc. Natl. Acad. Sci.  94:1857-1861

Turner, C.L. and A.K. Knapp.  1996.  Responses of a C₄ grass and three C₃ forbs to variation in nirogen and light in tallgrass prairie.  Ecology  77:1738-1749.

Williams, D.A., Jackson, L.L. and D.D. Smith.  2007.  Effects of frequent mowing on survival and persistence of forbs seeded into a species-poor grassland.  Restoration Ecology 15:24-33.

Wilsey, B.J. and H.W. Polley.  2002.  Plant species evenness reduces dicot invasion rates in Texas grasslands.  Ecology Letters 5:676-684.

Wilsey, B.J. and H.W. Polley.  2003.  Effects of seed additions and grazing history on diversity and aboveground productivity of sub-humid grasslands.  Ecology 84:920-931